It looks like the main factor to determine if someone has high mmr or not is simply the number of game as what they call 'experience' I think that the most you play after your first 1000 games you can improve your mmr without improving your real skills.
Basically it's just the fact that you play versus 'better player', in fact who has more games so losing or winning against them give you more mmr.
Advantage with this system is you may think you have improved when in fact not.
I think they should redo their system, first they should split their rankings and games with only solo queue players and games with only stacks can be 2 , 3 or 5 players. They should also do make a matchmaking mode with only solo queue players and where the matchmaking is random, in this matchmaking mode win ratio would mean something.
As I have explained in at least a dozen threads before... MMR is "true skill rating" type system. This means it has an uncertainty number in addition to a rating number. The uncertainty number controls how fast or slow your rating changes. When you win or lose as expected by the avg rankings of the teams your uncertainty decreases and the amount of points you can win or lose in the following game decreases. When you win or lose unexpectedly... the opposite of what the current avg tem ratings predicts... then your uncertainty increases and the value of your next game, win or lose, is bigger.
So if you go on a win streak or a loss streak the uncertainty increases and accelerates the change in your score. when you win/lose/win/lose in a consistent pattern your uncertainty becomes small and your rating stabilizes each game being worth fewer and fewer points.
Valve made in clear in the blog introducing this new ranked matching that the primary in-game diagnostic they are using to decide how to adjust your uncertainty and rating numbers besides win/loss is the integral of the difference in team gpm. You see this visually as the area between the line on the gpm graph and the zero-line. When your team is supposed to win (has higher avg MMRs) they expect your team to have a large gpm lead during most of the game. If you don't, win or lose, your score may go down. If you are expected to lose the game and you dominate gpm for most of the game, win or lose your score may go up. That is what it means. I'm not saying this is good or bad... just what it is.
It appears the Valve is using the same system for Ranked matches that they are for regular pubs so solo players who have high ratings will often be matched against a party since a full group of 10 players near their MMR does not exist in the que during a short 2 min perdio of time. If you want to get matches where everyone is very close to your MMR you can force the system to wait longer and have higher quality by cancelling the search at 30 seconds or 1 min or however long you want to wait. If a high quality match does not exist you may have to cancel it a re-search many times, but doing this will not let the search range expand long enough to put much lower MMR players on your team.
having a lot of trouble raising mine. on big wins where i feel like i play very well, it only goes up by like 10-50. with 5 wins in a row, i think i went up by a little less than 100. something interesting i discovered is that when i played heroes like jakiro and earthshaker, the game tracked how much long you keep people stunned/slowed. it reports this at the end of the game. http://i.imgur.com/AJ4YwKQ.jpg
true. on heroes like es/jakiro, they are skillshots though, so maybe that's something to consider. people should see what it reports for other heroes ^^
You all made the mistake of hope, remember this is the same matchmaking system. From the same people that brought you the muting system and language select. And it sounds like they did not even provide a clean slate, valve rating -1, uncertainty 0. Consistent to the end.
"Valve made in clear in the blog introducing this new ranked matching that the primary in-game diagnostic they are using to decide how to adjust your uncertainty and rating numbers besides win/loss is the integral of the difference in team gpm. You see this visually as the area between the line on the gpm graph and the zero-line. When your team is supposed to win (has higher avg MMRs) they expect your team to have a large gpm lead during most of the game. If you don't, win or lose, your score may go down. If you are expected to lose the game and you dominate gpm for most of the game, win or lose your score may go up. That is what it means. I'm not saying this is good or bad... just what it is."
I didn't read this part on any of their blog but actually it's quite interesting. It's sure that geting high gpm even when geting dominated is a signe of a good player the only issue I have with that is that i's kind of opposite to the objectives of the game which is geting to the thrones. Sometimes you may need to move into the map and try to get a gank to start a fight in 4 v 5 in high ground or to a t2 tower, now their system say that even when losing it's better to keep farming and dont care about the outcome of the game. Second point is that it advantage a lot stack players as they're more willing to stack camp or ancient for each others...
The issue with that is it advantage too much farming heroes, like luna, phantom lancer... these heroes can get crazy gpm; and also stupid support that will buy midas and not mekasnm at the early stage.
@Sampson
Well I said that it was a feeling. But basically what I feel is let's say a team true skill' would be 8/10 and an other team 8.5/10 but there would be so much team who are as good that they would be have a mmr of 3000 and 5000 and never play together.
I'd just recommend everyone chill about criticizing the rank they have until their uncertainty has reduces to a level that is known to be low. Everyone basically has a brand new rating in this mode (in terms of having high uncertainty).
As for the person asking why they must use a different rating: Well, if they didn't, it would contradict the credo to keep the hidden rating hidden.
Im currently at 4250, but does anyone know how much mmr you gain/lose per match is the first 10 matches? So far every game I've gotten has been pretty good/serious. I don't have that many games cause I just switched over in the beginning of summer so I'm kind of where I expect to be for solo que mmr.
I think it makes sense to base it off your normal matchmaking, I mean most people already has a lot of games clocked in there. It's way better than hon where if they just started everyone at 1500 it would be a complete crapshoot for a while if you win or lose.
@Revert Death. You have no idea about the weightings of different things. You have no idea how many points you get for winning or losing, or the effect that GPM/XPM, or K/D/A among a number of other factors have on a game. And no it does not benefit heroes that are capable of farming more because I can pretty much guarantee that your scores are compared to some kind of figure (the overall average, or the average within a particular MMR range, maybe two separate numbers for wins/losses, game duration, etc) achieved by other players with the SAME hero. Its like saying that heroes like Zeus/Spectre naturally get a higher rating because they get so many assists, which isn't the case when the hero itself is taken into account. An Anti-mage GPM/XPM or CS will not be compared to that of another hero, but compared against other results from the same hero.
You would probably find that taking a team fight and dying, but winning the match would yield a higher score than farming and farming and losing as a result.
Anyway, I think people are beginning to realise that the issue is match making itself. I can't really comment on the accuracy of the rating system because I can't just view anyone's rating, from what I have seen it seems alright - but match making. Oh boy..it seems like valve have decided that the average/total score of both teams being equal is ok, and do not take into account massive skill gaps between players in a match. The issue really comes down to this information not being available to players, so when noobs are playing a higher role (#1-2) and laning against someone that has like 2000 rating points higher, they don't know it, nor does their team, and they get stomped and you have a one sided game from very early on.
@Sampson
you say i am the elo i supost to be, ok let think about that.
I am in a low elo, so im this low elo i played 10 games and i won 8 of then so you really want to say i am supost to be on this elo???
and after a 8 streak win i just got +26points for a victory??? How do you expect me to get a high elo?? Since i got 3740 points and for each vitory i got about 26points, to be at 5500 i just need 92victory and 0 lost
And your system is so much greater? You list my friend here (http://dotabuff.com/players/57124770) with a higher rating than another friend here (http://dotabuff.com/players/86806476), when the first person I listed constantly asks for advice on how to proceed with the game, he also has difficulty when playing solo games, but when the 2nd person I listed uses the first guys account he has no trouble winning at all due to being placed in a lower mmr skill pool, so clearly valves system is working if it can deduce the appropriate pool of players
Can someone help make sense of this for me:
I've played 5 ranked games at the time of writing, won 3.
MMR dropped from 4075 to 3990 after the 5th game.
@Kwyjibo
Yeah it's hard to create a good matchmaking solution for a team game. I don't blame the guys at Valve, it's probably impossible to create a perfect system. It's really hard to quantify one's skill when there are 9 other people you have to take into consideration.
My criticism wasn't towards the scores themselves, more toward the fact that a really bad teammate ( I am talking 2 kills, 20 deaths, 3 assists guys ) can heavily degrade your ranked MMR just in one game. By amount that you'd have to win 4 games to recuperate it. Especially in the calibration phase when the scores are swinging wildly.
I must say though, I am pleasantly surprised that if you really play well and lose, you can gain a lot of rating (discarding those 1 in 50 cases when you've had a REALLY bad teammate ). This is a great improvement over a standard, flat Elo system. So there's a big plus to the MM guy at Valve for that! People won't feel as hopeless once they know they can go higher if they play well, regardless of circumstances ( Drunk teammates, people who have bad day, people sharing accounts etc. )
You dont get the meaning of a rating system, do you? I mean do you really cant self-relfect your skill and ignore the fact that you are not supposed to have 5,5k points? I mean there is a reason players like Swiftending etc. got such a high rating, because they are obviously a lot better than you. I dont understand as well why you cry about the fact that your "stuck" in a shit elo. It should be calculated based on your 700 games.
So if u wanna get out of your "shit elo" its pretty simple. Get more skilled. But to get more skill its normally takes longer than winning x amounts of games. Its a process which can take very long and often players reach their personal maximum pretty fast. I would say I play on the same skill level for about nearly 1-2 years and thats why Im "stuck" in my "elo-rating". But thats the nature of a ranking system. Not everyone can get to the very top, unless he becomes or is very skilled.
Nothing has changed, got tired of waiting for solo mode to be returned(if its going to be) and qued ranked. NOTHING has changed. Several random picks, English? not bloody likely(pun :>) a keeper who thinks hes a carry, void who ultis his own team more than enemy. Same ole shit, different mode.
PS. How do you see these shifts in your rank etc, is it only possible after the 10 games or is there a console command.
I haven't played a ranked solo matchmaking game yet, but I feel there is going to be a great disparity between the rating I get solo and the team rating I currently have of 4197. Every other person with a rating of 4K+ I see here plays either high or very high bracket consistently, while I usually play normal solo, and very high while queued with reasonably good friends. Not sure how I can be a better player while queued with friends, while solo I'm not there yet. Pretty odd. Will post later.
I still don't understand what's the difference between party and solo rating.
Doesn't the system just count your rating and your part members' rating, and then find a similar party with similar ratings?
My solo rating and party rating are close, thou, around 3750.
Solo rank just went up after my last game, but we lost. Thing is, when I queue in a party I'm playing very high and I feel the players are either as good or a little bit better than me, but in solo they're god-awful!
I lost 25 rating from this match after being hard support? Does this seem fair? Is there anything lacking in my statistics that could explain the system finding it fair that I lose rank? Genuine question, Im not trying to whine, Im just surprised I lost ranking in this game given my stats, my role, and the stats of the rest of the team.
@scrolawl, positive K/D/A doesn't mean you get a higher rating. What individual performance really means is that in losses, you will not lose as much rating as those who performed poorly, and if you did really, really well, only then would you actually end up gaining points.
^ Other players from your team dragged you down. If you get a feeder or two, your score WILL be lowered. It's an obvious flaw.
I've lost 100(!) points after just one game playing with Pudge who demanded mid and went 2/19/3.
If the game is close and you are doing fine, you won't lose rating and maybe even gain some. But if there is a really bad player in your team, he will drag everyone down.
It's really hard to create an algorithm that will behave properly in this situation. How your team fared, on average, is a really important metric and I believe it's one of the heavies weighted variables used by Valve's algorithm. I've also added some team-wide signals and it improved the accuracy of the rating significantly so I bet they also do that.
If thats how it works, that you will lose rank (almost) every time you lose, and similarily gain ranking every time you win, but that the amount you lose / gain is affected by how you performed, then that is balancing in the long run.
But so far Ive seen no indication that there is any influence of performance, only win / lose. I played a pretty careless razor game where I died needlessly many times and ended up with a pretty bad KDA compared to the rest of the team. Still gained ~20 points because we won, where Id be happier to see a smaller rank gain or even a loss in rank, because it would tell me the system takes performance into account.
The only game where there has been a difference from the +/- ~20 win/loss is a game where a guy rage quit when he got first blooded. That game only yielded +9 ranking when everyone left at around 6 min marl, instead of the normal ~20.
7 games into ranked mm and games are pretty similar to my normal mm. Well maybe people try a little harder... that's about it so far.
It looks like the main factor to determine if someone has high mmr or not is simply the number of game as what they call 'experience' I think that the most you play after your first 1000 games you can improve your mmr without improving your real skills.
Basically it's just the fact that you play versus 'better player', in fact who has more games so losing or winning against them give you more mmr.
Advantage with this system is you may think you have improved when in fact not.
I think they should redo their system, first they should split their rankings and games with only solo queue players and games with only stacks can be 2 , 3 or 5 players. They should also do make a matchmaking mode with only solo queue players and where the matchmaking is random, in this matchmaking mode win ratio would mean something.
Ranked matches are easier than usual pubs -_- http://dotabuff.com/matches/421432103
100% win so far
As I have explained in at least a dozen threads before... MMR is "true skill rating" type system. This means it has an uncertainty number in addition to a rating number. The uncertainty number controls how fast or slow your rating changes. When you win or lose as expected by the avg rankings of the teams your uncertainty decreases and the amount of points you can win or lose in the following game decreases. When you win or lose unexpectedly... the opposite of what the current avg tem ratings predicts... then your uncertainty increases and the value of your next game, win or lose, is bigger.
So if you go on a win streak or a loss streak the uncertainty increases and accelerates the change in your score. when you win/lose/win/lose in a consistent pattern your uncertainty becomes small and your rating stabilizes each game being worth fewer and fewer points.
Valve made in clear in the blog introducing this new ranked matching that the primary in-game diagnostic they are using to decide how to adjust your uncertainty and rating numbers besides win/loss is the integral of the difference in team gpm. You see this visually as the area between the line on the gpm graph and the zero-line. When your team is supposed to win (has higher avg MMRs) they expect your team to have a large gpm lead during most of the game. If you don't, win or lose, your score may go down. If you are expected to lose the game and you dominate gpm for most of the game, win or lose your score may go up. That is what it means. I'm not saying this is good or bad... just what it is.
------------------------------------------------------
Mostly likely what a 500 uncertainly literally means is 5X the base rate of change for your rating.
-------------------------------------------------------
It appears the Valve is using the same system for Ranked matches that they are for regular pubs so solo players who have high ratings will often be matched against a party since a full group of 10 players near their MMR does not exist in the que during a short 2 min perdio of time. If you want to get matches where everyone is very close to your MMR you can force the system to wait longer and have higher quality by cancelling the search at 30 seconds or 1 min or however long you want to wait. If a high quality match does not exist you may have to cancel it a re-search many times, but doing this will not let the search range expand long enough to put much lower MMR players on your team.
having a lot of trouble raising mine. on big wins where i feel like i play very well, it only goes up by like 10-50. with 5 wins in a row, i think i went up by a little less than 100. something interesting i discovered is that when i played heroes like jakiro and earthshaker, the game tracked how much long you keep people stunned/slowed. it reports this at the end of the game. http://i.imgur.com/AJ4YwKQ.jpg
That's an interesting factor, but very inaccurate... spamming stuns isn't better than using at the right time.
true. on heroes like es/jakiro, they are skillshots though, so maybe that's something to consider. people should see what it reports for other heroes ^^
Revert you misunderstood how the 'experience' function works......
Well, seeing Valve's rating in action I must say I anticipated something way better. Currently, their algorithm behaves absolutely retarded.
You can lose 100 points because one person decided to play pudge and his end game stats were 2/19/2.
Apparently, if the rest of your team was:
4/6/9
9/5/11
and you were 9/7/11
You deserve to lose 100 points :)
Working as intended.
I've anticipated something way better but I must say it sucks.
I am curious to see current top spots and their ratings.
You all made the mistake of hope, remember this is the same matchmaking system. From the same people that brought you the muting system and language select. And it sounds like they did not even provide a clean slate, valve rating -1, uncertainty 0. Consistent to the end.
why would it be a good idea to wipe away all of their previous data?
@ Vuccappella
the enemy team has alot more experience than us matches played so I doubt the difference was that great
I was playing OD and they tracked how much time i stunned someone, and by stunned i mean astral.
So yeah this was my first game, after this i got my 3300 mmr.
http://dotabuff.com/matches/421251880
@Relentless
"Valve made in clear in the blog introducing this new ranked matching that the primary in-game diagnostic they are using to decide how to adjust your uncertainty and rating numbers besides win/loss is the integral of the difference in team gpm. You see this visually as the area between the line on the gpm graph and the zero-line. When your team is supposed to win (has higher avg MMRs) they expect your team to have a large gpm lead during most of the game. If you don't, win or lose, your score may go down. If you are expected to lose the game and you dominate gpm for most of the game, win or lose your score may go up. That is what it means. I'm not saying this is good or bad... just what it is."
I didn't read this part on any of their blog but actually it's quite interesting. It's sure that geting high gpm even when geting dominated is a signe of a good player the only issue I have with that is that i's kind of opposite to the objectives of the game which is geting to the thrones. Sometimes you may need to move into the map and try to get a gank to start a fight in 4 v 5 in high ground or to a t2 tower, now their system say that even when losing it's better to keep farming and dont care about the outcome of the game. Second point is that it advantage a lot stack players as they're more willing to stack camp or ancient for each others...
The issue with that is it advantage too much farming heroes, like luna, phantom lancer... these heroes can get crazy gpm; and also stupid support that will buy midas and not mekasnm at the early stage.
@Sampson
Well I said that it was a feeling. But basically what I feel is let's say a team true skill' would be 8/10 and an other team 8.5/10 but there would be so much team who are as good that they would be have a mmr of 3000 and 5000 and never play together.
@Dorkly seems pretty insignificant then. :/
I'd just recommend everyone chill about criticizing the rank they have until their uncertainty has reduces to a level that is known to be low. Everyone basically has a brand new rating in this mode (in terms of having high uncertainty).
As for the person asking why they must use a different rating: Well, if they didn't, it would contradict the credo to keep the hidden rating hidden.
4k suck it bitches
Im currently at 4250, but does anyone know how much mmr you gain/lose per match is the first 10 matches? So far every game I've gotten has been pretty good/serious. I don't have that many games cause I just switched over in the beginning of summer so I'm kind of where I expect to be for solo que mmr.
I think it makes sense to base it off your normal matchmaking, I mean most people already has a lot of games clocked in there. It's way better than hon where if they just started everyone at 1500 it would be a complete crapshoot for a while if you win or lose.
Mine's around 3626
Won 10 games, lost 5.
Guess i'm the newest newbie around!
@Revert Death. You have no idea about the weightings of different things. You have no idea how many points you get for winning or losing, or the effect that GPM/XPM, or K/D/A among a number of other factors have on a game. And no it does not benefit heroes that are capable of farming more because I can pretty much guarantee that your scores are compared to some kind of figure (the overall average, or the average within a particular MMR range, maybe two separate numbers for wins/losses, game duration, etc) achieved by other players with the SAME hero. Its like saying that heroes like Zeus/Spectre naturally get a higher rating because they get so many assists, which isn't the case when the hero itself is taken into account. An Anti-mage GPM/XPM or CS will not be compared to that of another hero, but compared against other results from the same hero.
You would probably find that taking a team fight and dying, but winning the match would yield a higher score than farming and farming and losing as a result.
Anyway, I think people are beginning to realise that the issue is match making itself. I can't really comment on the accuracy of the rating system because I can't just view anyone's rating, from what I have seen it seems alright - but match making. Oh boy..it seems like valve have decided that the average/total score of both teams being equal is ok, and do not take into account massive skill gaps between players in a match. The issue really comes down to this information not being available to players, so when noobs are playing a higher role (#1-2) and laning against someone that has like 2000 rating points higher, they don't know it, nor does their team, and they get stomped and you have a one sided game from very early on.
@Sampson
you say i am the elo i supost to be, ok let think about that.
I am in a low elo, so im this low elo i played 10 games and i won 8 of then so you really want to say i am supost to be on this elo???
and after a 8 streak win i just got +26points for a victory??? How do you expect me to get a high elo?? Since i got 3740 points and for each vitory i got about 26points, to be at 5500 i just need 92victory and 0 lost
@saboia, go play checkers or something, you are a complete retard
2736 Solo
Won 8/11 games
@terrible
explain why?
i want to understand
@Fearohnum
And your system is so much greater? You list my friend here (http://dotabuff.com/players/57124770) with a higher rating than another friend here (http://dotabuff.com/players/86806476), when the first person I listed constantly asks for advice on how to proceed with the game, he also has difficulty when playing solo games, but when the 2nd person I listed uses the first guys account he has no trouble winning at all due to being placed in a lower mmr skill pool, so clearly valves system is working if it can deduce the appropriate pool of players
Can someone help make sense of this for me:
I've played 5 ranked games at the time of writing, won 3.
MMR dropped from 4075 to 3990 after the 5th game.
My 2 losses carry more weightage than my 3 wins?
4970, but only played 1 ranked so far.
@Kwyjibo
Yeah it's hard to create a good matchmaking solution for a team game. I don't blame the guys at Valve, it's probably impossible to create a perfect system. It's really hard to quantify one's skill when there are 9 other people you have to take into consideration.
My criticism wasn't towards the scores themselves, more toward the fact that a really bad teammate ( I am talking 2 kills, 20 deaths, 3 assists guys ) can heavily degrade your ranked MMR just in one game. By amount that you'd have to win 4 games to recuperate it. Especially in the calibration phase when the scores are swinging wildly.
I must say though, I am pleasantly surprised that if you really play well and lose, you can gain a lot of rating (discarding those 1 in 50 cases when you've had a REALLY bad teammate ). This is a great improvement over a standard, flat Elo system. So there's a big plus to the MM guy at Valve for that! People won't feel as hopeless once they know they can go higher if they play well, regardless of circumstances ( Drunk teammates, people who have bad day, people sharing accounts etc. )
@saboia
You dont get the meaning of a rating system, do you? I mean do you really cant self-relfect your skill and ignore the fact that you are not supposed to have 5,5k points? I mean there is a reason players like Swiftending etc. got such a high rating, because they are obviously a lot better than you. I dont understand as well why you cry about the fact that your "stuck" in a shit elo. It should be calculated based on your 700 games.
So if u wanna get out of your "shit elo" its pretty simple. Get more skilled. But to get more skill its normally takes longer than winning x amounts of games. Its a process which can take very long and often players reach their personal maximum pretty fast. I would say I play on the same skill level for about nearly 1-2 years and thats why Im "stuck" in my "elo-rating". But thats the nature of a ranking system. Not everyone can get to the very top, unless he becomes or is very skilled.
stackers gonna brag about solo rating
That doesn't even make sense, rating for both solo and parties are tracked separately.
Nothing has changed, got tired of waiting for solo mode to be returned(if its going to be) and qued ranked. NOTHING has changed. Several random picks, English? not bloody likely(pun :>) a keeper who thinks hes a carry, void who ultis his own team more than enemy. Same ole shit, different mode.
PS. How do you see these shifts in your rank etc, is it only possible after the 10 games or is there a console command.
TY Trillion.
I haven't played a ranked solo matchmaking game yet, but I feel there is going to be a great disparity between the rating I get solo and the team rating I currently have of 4197. Every other person with a rating of 4K+ I see here plays either high or very high bracket consistently, while I usually play normal solo, and very high while queued with reasonably good friends. Not sure how I can be a better player while queued with friends, while solo I'm not there yet. Pretty odd. Will post later.
Silence, there is a console command to see your current ranking:
developer 1
dota_game_account_debug
4202
i feel its pretty average
Well, I was right. Massive disparity.
competitive_rank: 4197
competitive_rank_uncertainty: 901
calibration_games_remaining: 9
solo_competitive_rank: 3406
solo_competitive_rank_uncertainty: 946
solo_calibration_games_remaining: 9
How is this even possible? :S
I still don't understand what's the difference between party and solo rating.
Doesn't the system just count your rating and your part members' rating, and then find a similar party with similar ratings?
My solo rating and party rating are close, thou, around 3750.
5495 with 5-5 (solo). 5 of these games were against/with Dendi though. Some of them had other pros as well. Also what's the highest number anyway?
^
if im not mistaken around 6,3k is the highest
5200 solo (1 game)
Took 25 minutes to queue for my first ranked game.....
competitive_rank: 4197
competitive_rank_uncertainty: 901
calibration_games_remaining: 9
solo_competitive_rank: 3532
solo_competitive_rank_uncertainty: 913
Solo rank just went up after my last game, but we lost. Thing is, when I queue in a party I'm playing very high and I feel the players are either as good or a little bit better than me, but in solo they're god-awful!
Im at around 3900 rating.
Seems to be a lot of higher rated players in this thread. What do you all think about this game?
http://dotabuff.com/matches/421711926
I lost 25 rating from this match after being hard support? Does this seem fair? Is there anything lacking in my statistics that could explain the system finding it fair that I lose rank? Genuine question, Im not trying to whine, Im just surprised I lost ranking in this game given my stats, my role, and the stats of the rest of the team.
@scrolawl, positive K/D/A doesn't mean you get a higher rating. What individual performance really means is that in losses, you will not lose as much rating as those who performed poorly, and if you did really, really well, only then would you actually end up gaining points.
^ Other players from your team dragged you down. If you get a feeder or two, your score WILL be lowered. It's an obvious flaw.
I've lost 100(!) points after just one game playing with Pudge who demanded mid and went 2/19/3.
If the game is close and you are doing fine, you won't lose rating and maybe even gain some. But if there is a really bad player in your team, he will drag everyone down.
It's really hard to create an algorithm that will behave properly in this situation. How your team fared, on average, is a really important metric and I believe it's one of the heavies weighted variables used by Valve's algorithm. I've also added some team-wide signals and it improved the accuracy of the rating significantly so I bet they also do that.
Any chance for some kind of a ladder to Dotabuff? Or can't these ratings be seen by API?
@Terrible.
If thats how it works, that you will lose rank (almost) every time you lose, and similarily gain ranking every time you win, but that the amount you lose / gain is affected by how you performed, then that is balancing in the long run.
But so far Ive seen no indication that there is any influence of performance, only win / lose. I played a pretty careless razor game where I died needlessly many times and ended up with a pretty bad KDA compared to the rest of the team. Still gained ~20 points because we won, where Id be happier to see a smaller rank gain or even a loss in rank, because it would tell me the system takes performance into account.
The only game where there has been a difference from the +/- ~20 win/loss is a game where a guy rage quit when he got first blooded. That game only yielded +9 ranking when everyone left at around 6 min marl, instead of the normal ~20.
this ranked mm is basically the same mm we had
thanks gabe you fucking idiot
^lold
Anyone else notice a disparity between their solo and party ranked MMR figures?